import os
import scipy
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns

from sklearn.model_selection import train_test_split
from sklearn.linear_model import LinearRegression

import numpy as np
 
from sklearn.model_selection import cross_validate
from sklearn.model_selection import cross_val_score
from sklearn.metrics import mean_squared_error
from sklearn.metrics import r2_score
from sklearn.compose import make_column_selector as selector
from sklearn.preprocessing import  OneHotEncoder, StandardScaler, PolynomialFeatures
from sklearn.ensemble import RandomForestRegressor
from sklearn.pipeline import Pipeline
from sklearn.tree import DecisionTreeRegressor
from sklearn.model_selection import GridSearchCV
from time import time
from sklearn.compose import ColumnTransformer
from scipy.stats import norm
import scipy.stats as stats
import numpy as np

_DATA_PATH = 'data/EEHPA.csv'
_SIHTTUNNUS_ = 'BR'
_GOAL_ = 'Leida parima mitme argumendiga regressiooni mudeli sihttunnuse BR ennustamise jaoks.'
_DROP_= ['age', 'weight', 'height', 'bmi', 'ID',
         'gaAnkle', 'gaChest', 'gaWrist', 'equivital', 'cosmed', 'COSMEDset_row',
         'trial_date',
         'VE.VO2', 'VE.VCO2', 'R', 'FiO2', 'FiCO2',
         'VO2', 'VCO2', 'EEm', 'EEh',
         'Ti', 'Te', 'Ttot',
         'Qt', 'SV',
         'predicted_activity_label']
_OBJ_CAST_= []
_DROP_UNNAMED_ = True

df = pd.read_csv(_DATA_PATH)
df

print(f"Andmestikus on \033[1m{df.shape[0]}\033[0m ridu, neid iseloomustab \033[1m{df.shape[1]}\033[0m tunnust.")

Andmestikus on 38723 ridu, neid iseloomustab 41 tunnust.

df.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 38723 entries, 0 to 38722
Data columns (total 41 columns):
 #   Column                    Non-Null Count  Dtype  
---  ------                    --------------  -----  
 0   ID                        38723 non-null  object 
 1   trial_date                38723 non-null  object 
 2   gender                    38723 non-null  object 
 3   age                       38723 non-null  float64
 4   weight                    38723 non-null  float64
 5   height                    38723 non-null  int64  
 6   bmi                       38723 non-null  float64
 7   gaAnkle                   38723 non-null  bool   
 8   gaChest                   38723 non-null  bool   
 9   gaWrist                   38723 non-null  bool   
 10  equivital                 38723 non-null  bool   
 11  cosmed                    38723 non-null  bool   
 12  EEm                       38723 non-null  float64
 13  COSMEDset_row             38723 non-null  int64  
 14  EEh                       38723 non-null  float64
 15  EEtot                     38723 non-null  float64
 16  METS                      38723 non-null  float64
 17  Rf                        38723 non-null  float64
 18  BR                        38723 non-null  int64  
 19  VT                        38723 non-null  float64
 20  VE                        38723 non-null  float64
 21  VO2                       38723 non-null  float64
 22  VCO2                      38723 non-null  float64
 23  O2exp                     38723 non-null  float64
 24  CO2exp                    38723 non-null  float64
 25  FeO2                      38723 non-null  float64
 26  FeCO2                     38723 non-null  float64
 27  FiO2                      38723 non-null  float64
 28  FiCO2                     38723 non-null  float64
 29  VE.VO2                    38723 non-null  float64
 30  VE.VCO2                   38723 non-null  float64
 31  R                         38723 non-null  float64
 32  Ti                        38723 non-null  float64
 33  Te                        38723 non-null  float64
 34  Ttot                      38723 non-null  float64
 35  VO2.HR                    38723 non-null  float64
 36  HR                        38723 non-null  int64  
 37  Qt                        38723 non-null  int64  
 38  SV                        38723 non-null  int64  
 39  original_activity_labels  24452 non-null  object 
 40  predicted_activity_label  11395 non-null  object 
dtypes: bool(5), float64(25), int64(6), object(5)
memory usage: 10.8+ MB

print_prop_num_count = len(df.select_dtypes(exclude=object).columns)
print_prop_obj_count = len(df.select_dtypes(include=object).columns)

print(f"Andmestikus on \033[1m{print_prop_num_count}\033[0m arvulist ja \033[1m{print_prop_obj_count}\033[0m mittearvulist muutujat")

Andmestikus on 36 arvulist ja 5 mittearvulist muutujat

if len(_OBJ_CAST_) > 0:
    obj_cast_l = {}
    for i in _OBJ_CAST_:
        if i in df:
            obj_cast_l.update({i: str})
    
    df = df.astype(obj_cast_l)
        
    df.info()

if len(_OBJ_CAST_) > 0:
    
    print("Teisendame järgmised tunnused kategoriaalseteks:")
    print(f"\033[1m{[i for i in _OBJ_CAST_]}\033[0m")
    
    print_prop_num_count = len(df.select_dtypes(exclude=object).columns)
    print_prop_obj_count = len(df.select_dtypes(include=object).columns)

    print(f"Andmestikus on \033[1m{print_prop_num_count}\033[0m arvulist ja \033[1m{print_prop_obj_count}\033[0m mittearvulist muutujat")

duplicates = df.duplicated(keep='first').sum();
print(f"Andmestikus on leitud {duplicates} duplikaate.")

Andmestikus on leitud 28 duplikaate.

df.drop_duplicates(keep='first',inplace=True)
duplicates_new = df.duplicated(keep='first').sum()
print(f"Peale puhastamist andmestikus on leitud {duplicates_new} duplikaate.")

Peale puhastamist andmestikus on leitud 0 duplikaate.

missing_values_validation = df.isna().sum()
missing_values_validation

ID                              0
trial_date                      0
gender                          0
age                             0
weight                          0
height                          0
bmi                             0
gaAnkle                         0
gaChest                         0
gaWrist                         0
equivital                       0
cosmed                          0
EEm                             0
COSMEDset_row                   0
EEh                             0
EEtot                           0
METS                            0
Rf                              0
BR                              0
VT                              0
VE                              0
VO2                             0
VCO2                            0
O2exp                           0
CO2exp                          0
FeO2                            0
FeCO2                           0
FiO2                            0
FiCO2                           0
VE.VO2                          0
VE.VCO2                         0
R                               0
Ti                              0
Te                              0
Ttot                            0
VO2.HR                          0
HR                              0
Qt                              0
SV                              0
original_activity_labels    14263
predicted_activity_label    27328
dtype: int64

print(f"Puhastatud andmestikus on \033[1m{df.shape[0]}\033[0m ridu, neid iseloomustab \033[1m{df.shape[1]}\033[0m tunnust.")

Puhastatud andmestikus on 38695 ridu, neid iseloomustab 41 tunnust.

if len(_DROP_) > 0:
    prep_validated_drop_list = []
    for i in _DROP_:
        if i in df:
            prep_validated_drop_list.append(i)
    
    if len(prep_validated_drop_list) > 0:
        df.drop(columns=prep_validated_drop_list,axis = 1, inplace = True)

if _DROP_UNNAMED_:   
    prep_empty_cols = df.columns.str.contains('unnamed',case = False)

    if len(np.where(prep_empty_cols == True)) > 0:
        df.drop(df.columns[prep_empty_cols],axis = 1, inplace = True)
    
df

print_prop_num_count = len(df.select_dtypes(exclude=object).columns)
print_prop_obj_count = len(df.select_dtypes(include=object).columns)

print(f"Andmestikus on \033[1m{print_prop_num_count}\033[0m arvulist ja \033[1m{print_prop_obj_count}\033[0m mittearvulist muutujat")

Andmestikus on 12 arvulist ja 2 mittearvulist muutujat

pd.set_option('display.float_format', lambda x: '%.2f' % x)
df.describe().T

feature_columns=df.drop(_SIHTTUNNUS_, axis=1).select_dtypes(exclude=object).columns
feature_columns

Index(['EEtot', 'METS', 'Rf', 'VT', 'VE', 'O2exp', 'CO2exp', 'FeO2', 'FeCO2',
       'VO2.HR', 'HR'],
      dtype='object')

fig, axs = plt.subplots(len(feature_columns),2,dpi=95,figsize=(15,30))
i = 0
for col in feature_columns:
    df[col].plot(kind='hist',ax=axs[i,0], title=col, color="steelblue")
    df[col].plot(kind='box',vert=False,ax=axs[i,1], title=col,
                                             patch_artist = True,
           boxprops = dict(facecolor = "steelblue"),
                                             medianprops = dict(color = "red", linewidth = 1.5)).set_yticklabels('')

    i+=1
fig.tight_layout()
plt.show()

df.describe(include=[object]).T

for column in df.select_dtypes(include=object).columns:
    print(column)
    print(df[column].value_counts().sort_index())
    print()

gender
gender
female    13909
male      24786
Name: count, dtype: int64

original_activity_labels
original_activity_labels
cycling            4212
dishwashing        1885
lyingDownLeft      1459
lyingDownRight     1307
sittingChair       1546
sittingCouch       1623
sittingSofa        1549
stakingShelves     1689
standing           1102
step                413
syncJumping         161
vacuumCleaning     1744
walkingFast        1997
walkingNormal      1883
walkingSlow        1694
walkingStairsUp     168
Name: count, dtype: int64

print(f"Sihttunnus: Y = \033[1m{_SIHTTUNNUS_}\033[0m.")

Sihttunnus: Y = BR.

fig, axs = plt.subplots(1,2,dpi=95,figsize=(15,5))
df[_SIHTTUNNUS_].plot(kind='hist',ax=axs[0], title="{}".format(_SIHTTUNNUS_), color="steelblue")
df[_SIHTTUNNUS_].plot(kind='box',vert=False,ax=axs[1], title="{}".format(_SIHTTUNNUS_),
                                             patch_artist = True,
           boxprops = dict(facecolor = "steelblue"),
                                             medianprops = dict(color = "red", linewidth = 1.5))
plt.show()

sns.pairplot(df.select_dtypes(exclude=object))
plt.show()

df.select_dtypes(exclude=object).corr()

num_f=df.select_dtypes(exclude=object)

plt.figure(figsize=(16,16))
sns.heatmap(num_f.corr(), annot=True, fmt= '.2f')
plt.show()

plt.figure(figsize=(16,16))
sns.set(font_scale=1.0)
hm = sns.heatmap(num_f.corr(), 
                 cbar=True, 
                 annot=True, 
                 square=True, 
                 fmt='.2f',
                 annot_kws={'size': 10}, 
                 yticklabels=num_f.columns,
                 xticklabels=num_f.columns,
                 cmap=sns.diverging_palette(10, 220, sep=30, n=256),
                 center=0.0)
plt.show()

sns.set(font_scale=1.0)
km = sns.clustermap(num_f.corr(), 
                    cbar=True, 
                    annot=True,
                    fmt='.2f',
                    annot_kws={'size': 10}, 
                    yticklabels=num_f.columns,
                    xticklabels=num_f.columns,
                    cmap=sns.diverging_palette(10, 220, sep=30, n=256),
                    center=0.0)
plt.show()

pd.set_option('display.float_format', lambda x: '%.4f' % x)
num_f.corrwith(df[_SIHTTUNNUS_]).sort_values()

VE       -0.9704
METS     -0.9097
O2exp    -0.7376
VT       -0.7272
CO2exp   -0.6652
VO2.HR   -0.6300
EEtot    -0.5134
HR       -0.4657
Rf       -0.3460
FeCO2    -0.3060
FeO2      0.1410
BR        1.0000
dtype: float64

plt.figure(dpi=130,figsize=(1,4))
sns.set(font_scale=0.8)
sns.heatmap(pd.DataFrame(num_f.corrwith(df[_SIHTTUNNUS_]).sort_values()), fmt='.2f',
            annot=True, cmap=sns.diverging_palette(10, 220, sep=30, n=256),
                 center=0.0)
plt.show()

categ_columns=df.select_dtypes(include=object).columns
fig, axs = plt.subplots(len(categ_columns),1,dpi=95,figsize=(15,25))
i = 0
for col in categ_columns:
        df.boxplot(
            column=[_SIHTTUNNUS_],
            by=col,
            ax=axs[i],
            patch_artist = True,
            boxprops = dict(facecolor = "steelblue"),
            medianprops = dict(color = "red", linewidth = 1.5)
        )
        i+=1
plt.suptitle('')
fig.tight_layout()
plt.show()

pd.set_option('display.float_format', lambda x: '%.2f' % x)
for col in categ_columns:
    print(df.groupby(col)[_SIHTTUNNUS_].describe())
    print()

          count  mean   std   min   25%   50%   75%   max
gender                                                   
female 13909.00 82.22 10.11 34.00 78.00 85.00 90.00 99.00
male   24786.00 81.90 10.58 23.00 77.00 85.00 90.00 99.00

                           count  mean   std   min   25%   50%   75%   max
original_activity_labels                                                  
cycling                  4212.00 64.75 11.82 23.00 57.00 64.00 73.00 99.00
dishwashing              1885.00 86.46  3.90 63.00 84.00 87.00 89.00 98.00
lyingDownLeft            1459.00 86.42  5.25 53.00 84.00 87.00 90.00 99.00
lyingDownRight           1307.00 90.19  3.40 62.00 89.00 91.00 92.00 98.00
sittingChair             1546.00 90.25  3.33 61.00 89.00 91.00 92.00 99.00
sittingCouch             1623.00 90.28  3.15 69.00 89.00 91.00 92.00 98.00
sittingSofa              1549.00 90.32  3.65 63.00 89.00 91.00 93.00 98.00
stakingShelves           1689.00 85.42  4.80 62.00 83.00 86.00 89.00 98.00
standing                 1102.00 83.45  5.76 61.00 80.00 84.00 88.00 99.00
step                      413.00 83.70  4.94 68.00 81.00 83.00 87.00 97.00
syncJumping               161.00 86.94  6.27 61.00 85.00 88.00 91.00 98.00
vacuumCleaning           1744.00 82.10  5.83 49.00 78.00 83.00 86.00 99.00
walkingFast              1997.00 72.11  8.75 45.00 67.00 72.00 77.00 99.00
walkingNormal            1883.00 76.98  7.32 43.00 72.00 77.00 82.00 98.00
walkingSlow              1694.00 82.11  5.78 62.00 78.00 82.00 86.00 98.00
walkingStairsUp           168.00 87.27  5.12 62.00 84.75 88.00 91.00 97.00

num_f=df.select_dtypes(exclude=object)
X = num_f.drop([_SIHTTUNNUS_],axis=1)
y = num_f[_SIHTTUNNUS_]

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=0)

pd.set_option('display.float_format', lambda x: '%.4f' % x)
from sklearn.preprocessing import StandardScaler
sc = StandardScaler()
sc.fit(X_train)
X_train_std = sc.transform(X_train)
X_test_std = sc.transform(X_test)

std_df = pd.DataFrame(X_train_std, columns=X.columns)
std_df.describe().T

slr = LinearRegression()

slr.fit(X_train_std, y_train)

LinearRegression()

LinearRegression()

slr.coef_

array([ 0.21141921, -4.86764595, -0.16945791,  4.94370014, -6.65009999,
       -7.21399124,  3.17396404, -0.77187586, -1.11605182,  1.19794119,
       -0.43799687])

for col_name, x_i in zip(X.columns, slr.coef_):
    print(col_name  + "\t", round(x_i, 4))

EEtot	 0.2114
METS	 -4.8676
Rf	 -0.1695
VT	 4.9437
VE	 -6.6501
O2exp	 -7.214
CO2exp	 3.174
FeO2	 -0.7719
FeCO2	 -1.1161
VO2.HR	 1.1979
HR	 -0.438

coefs = pd.DataFrame(slr.coef_, columns=["Coefficients"], index=X.columns)
coefs

coefs.plot(kind="barh", figsize=(9, 7))
plt.title("MLR model")
plt.axvline(x=0, color=".5")
plt.subplots_adjust(left=0.3)

cv_model = cross_validate(
    slr,
    X_train_std,
    y_train,
    cv=10,
    n_jobs=1
)
coefs = pd.DataFrame(
    [slr.coef_ for model in cv_model],
    columns=X.columns,
)
plt.figure(figsize=(9, 7))
sns.boxplot(data=coefs, orient="h", color="cyan", saturation=0.5)
plt.axvline(x=0, color=".5")
plt.xlabel("Coefficient importance")
plt.title("Coefficient importance and its variability")
plt.subplots_adjust(left=0.3)

scores = cross_val_score(estimator=slr,
                         X=X_train_std,
                         y=y_train,
                         cv=10,
                         n_jobs=1)
print('CV keskmine R2 täpsus: %.3f' % np.mean(scores), "+/- %.3f" % np.std(scores))
print('R2 täpsus treeningandmetel: %.3f' % slr.score(X_train_std, y_train))
print('R2 täpsus testandmetel: %.3f' % slr.score(X_test_std, y_test))

CV keskmine R2 täpsus: 0.970 +/- 0.001
R2 täpsus treeningandmetel: 0.970
R2 täpsus testandmetel: 0.970

scores = cross_val_score(estimator=slr,
                         X=X_train_std,
                         y=y_train,
                         scoring = 'neg_mean_squared_error',
                         cv=10,
                         n_jobs=1)
print('CV keskmine RMSE: %.3f' % np.mean(np.sqrt(np.abs(scores))), "+/- %.3f" % np.std(np.sqrt(np.abs(scores))))
print('RMSE treeningandmetel: %.3f' % np.sqrt(mean_squared_error(y_train,slr.predict(X_train_std))))
print('RMSE testandmetel: %.3f' % np.sqrt(mean_squared_error(y_test,slr.predict(X_test_std))))

CV keskmine RMSE: 1.808 +/- 0.042
RMSE treeningandmetel: 1.807
RMSE testandmetel: 1.786

residuals=y_train-slr.predict(X_train_std)

std_residuals=residuals/np.std(residuals)

fig, axs = plt.subplots(2,2,dpi=95,figsize=(15,15))
plt.style.use("seaborn-v0_8-whitegrid")
# Residual against fitted values
axs[0, 0].scatter(x=slr.predict(X_train_std), y=std_residuals)
axs[0, 0].axhline(y=0, color='red', linestyle='dashed')
axs[0, 0].set_xlabel('Fitted Values')
axs[0, 0].set_ylabel('Std. Residuals')
axs[0, 0].set_title('Residuals vs Fitted')

# normal qqplot
stats.probplot(std_residuals, plot=axs[0, 1])
#sm.qqplot(std_residuals, dist=stats.t, fit=True, line='45', c='#4C72B0',ax=axs[0, 1])
axs[0, 1].set_title('Normal Q-Q')

# Fitted values against actual values
axs[1, 0].scatter(x=y_train, y=slr.predict(X_train_std))
axs[1, 0].plot(y_train, y_train, color='red', linestyle='dashed')
axs[1, 0].set_xlabel('Actual Values')
axs[1, 0].set_ylabel('Fitted Values')
axs[1, 0].set_title('Fitted vs Actual')

# Histogram of std. residuals
axs[1, 1].hist(std_residuals, density=True)
x = np.linspace(min(std_residuals),max(std_residuals), 500)
axs[1, 1].plot(x, norm.pdf(x),color='red')
axs[1, 1].set_xlabel('Std. Residuals')
axs[1, 1].set_title('Std. Residuals Density Plot')
fig.tight_layout()
plt.show()

X_dummy = pd.get_dummies(data=df.drop([_SIHTTUNNUS_],axis=1), drop_first=True)
X_dummy.head

<bound method NDFrame.head of         EEtot   METS      Rf     VT      VE    O2exp  CO2exp    FeO2  FeCO2  \
0      0.0000 0.9528 21.5054 0.4091  8.7978  73.1476 12.4463 17.8803 3.0424   
1      0.0097 1.0143 24.5902 0.4642 11.4144  85.4919 11.8342 18.4176 2.5495   
2      0.0940 1.9223 22.5564 0.8774 19.7902 159.3537 25.3007 18.1628 2.8837   
3      0.2080 2.0189 20.0000 0.9243 18.4858 164.5567 30.6057 17.8035 3.3113   
4      0.2703 1.2154 22.0588 0.5876 12.9624 107.3240 16.2210 18.2639 2.7604   
...       ...    ...     ...    ...     ...      ...     ...     ...    ...   
38718 85.4218 1.2031 21.8182 0.5650 12.3281 102.0947 13.8710 18.0686 2.4549   
38719 85.4871 0.9703 19.8675 0.5314 10.5573  96.9951 11.8312 18.2534 2.2265   
38720 85.5828 2.2391 26.7857 0.6436 17.2386 110.4581 19.6846 17.1632 3.0586   
38721 85.7260 1.3556 14.9254 0.7017 10.4733 120.3514 22.2344 17.1512 3.1686   
38722 86.1708 1.4569 24.7934 0.6028 14.9449 109.3362 12.8511 18.1387 2.1320   

       VO2.HR  ...  original_activity_labels_sittingSofa  \
0      2.1250  ...                                 False   
1      2.2403  ...                                 False   
2      4.2051  ...                                 False   
3      4.3329  ...                                 False   
4      2.6086  ...                                 False   
...       ...  ...                                   ...   
38718  3.9375  ...                                 False   
38719  3.1757  ...                                 False   
38720  7.3279  ...                                 False   
38721  4.4365  ...                                 False   
38722  4.7681  ...                                 False   

       original_activity_labels_stakingShelves  \
0                                        False   
1                                        False   
2                                        False   
3                                        False   
4                                        False   
...                                        ...   
38718                                    False   
38719                                    False   
38720                                    False   
38721                                    False   
38722                                    False   

       original_activity_labels_standing  original_activity_labels_step  \
0                                  False                          False   
1                                  False                          False   
2                                  False                          False   
3                                  False                          False   
4                                  False                          False   
...                                  ...                            ...   
38718                              False                          False   
38719                              False                          False   
38720                              False                          False   
38721                              False                          False   
38722                              False                          False   

       original_activity_labels_syncJumping  \
0                                     False   
1                                     False   
2                                     False   
3                                     False   
4                                     False   
...                                     ...   
38718                                 False   
38719                                 False   
38720                                 False   
38721                                 False   
38722                                 False   

       original_activity_labels_vacuumCleaning  \
0                                        False   
1                                        False   
2                                        False   
3                                        False   
4                                        False   
...                                        ...   
38718                                    False   
38719                                    False   
38720                                    False   
38721                                    False   
38722                                    False   

       original_activity_labels_walkingFast  \
0                                     False   
1                                     False   
2                                     False   
3                                     False   
4                                     False   
...                                     ...   
38718                                 False   
38719                                 False   
38720                                 False   
38721                                 False   
38722                                 False   

       original_activity_labels_walkingNormal  \
0                                       False   
1                                       False   
2                                       False   
3                                       False   
4                                       False   
...                                       ...   
38718                                   False   
38719                                   False   
38720                                   False   
38721                                   False   
38722                                   False   

       original_activity_labels_walkingSlow  \
0                                     False   
1                                     False   
2                                     False   
3                                     False   
4                                     False   
...                                     ...   
38718                                 False   
38719                                 False   
38720                                 False   
38721                                 False   
38722                                 False   

       original_activity_labels_walkingStairsUp  
0                                         False  
1                                         False  
2                                         False  
3                                         False  
4                                         False  
...                                         ...  
38718                                     False  
38719                                     False  
38720                                     False  
38721                                     False  
38722                                     False  

[38695 rows x 27 columns]>

X_dummy

X_train, X_test, y_train, y_test = train_test_split(X_dummy, y, test_size=0.2, random_state=0)

X_train_num = X_train[num_f.drop([_SIHTTUNNUS_],axis=1).columns]
X_train_num.head()

X_test_num = X_test[num_f.drop([_SIHTTUNNUS_],axis=1).columns]
X_test_num.head()

sc.fit(X_train_num)
X_train_std = sc.transform(X_train_num)
X_test_std = sc.transform(X_test_num)

X_train_num.columns

Index(['EEtot', 'METS', 'Rf', 'VT', 'VE', 'O2exp', 'CO2exp', 'FeO2', 'FeCO2',
       'VO2.HR', 'HR'],
      dtype='object')

X_train.columns

Index(['EEtot', 'METS', 'Rf', 'VT', 'VE', 'O2exp', 'CO2exp', 'FeO2', 'FeCO2',
       'VO2.HR', 'HR', 'gender_male', 'original_activity_labels_dishwashing',
       'original_activity_labels_lyingDownLeft',
       'original_activity_labels_lyingDownRight',
       'original_activity_labels_sittingChair',
       'original_activity_labels_sittingCouch',
       'original_activity_labels_sittingSofa',
       'original_activity_labels_stakingShelves',
       'original_activity_labels_standing', 'original_activity_labels_step',
       'original_activity_labels_syncJumping',
       'original_activity_labels_vacuumCleaning',
       'original_activity_labels_walkingFast',
       'original_activity_labels_walkingNormal',
       'original_activity_labels_walkingSlow',
       'original_activity_labels_walkingStairsUp'],
      dtype='object')

dummy_col=X_train.columns[~X_train.columns.isin(X_train_num.columns)]
dummy_col

Index(['gender_male', 'original_activity_labels_dishwashing',
       'original_activity_labels_lyingDownLeft',
       'original_activity_labels_lyingDownRight',
       'original_activity_labels_sittingChair',
       'original_activity_labels_sittingCouch',
       'original_activity_labels_sittingSofa',
       'original_activity_labels_stakingShelves',
       'original_activity_labels_standing', 'original_activity_labels_step',
       'original_activity_labels_syncJumping',
       'original_activity_labels_vacuumCleaning',
       'original_activity_labels_walkingFast',
       'original_activity_labels_walkingNormal',
       'original_activity_labels_walkingSlow',
       'original_activity_labels_walkingStairsUp'],
      dtype='object')

X_train_std=pd.DataFrame(X_train_std, columns=X_train_num.columns).join(X_train[dummy_col].reset_index()).drop(['index'],axis=1)
X_train_std.head()

X_test_std=pd.DataFrame(X_test_std, columns=X_test_num.columns).join(X_test[dummy_col].reset_index()).drop(['index'],axis=1)
X_test_std.head()

X_train_std.describe().T

slr = LinearRegression()

slr.fit(X_train_std, y_train)

LinearRegression()

LinearRegression()

for col_name, x_i in zip(X_train_std.columns, slr.coef_):
    print(col_name  + "\t", round(x_i, 4))

EEtot	 0.3412
METS	 -3.598
Rf	 -0.1433
VT	 -1.9545
VE	 -7.7387
O2exp	 -1.0557
CO2exp	 3.7042
FeO2	 -0.9021
FeCO2	 -1.1345
VO2.HR	 1.0252
HR	 -0.431
gender_male	 1.6887
original_activity_labels_dishwashing	 -0.031
original_activity_labels_lyingDownLeft	 -0.0819
original_activity_labels_lyingDownRight	 0.1211
original_activity_labels_sittingChair	 0.1296
original_activity_labels_sittingCouch	 0.2812
original_activity_labels_sittingSofa	 0.3217
original_activity_labels_stakingShelves	 -0.0936
original_activity_labels_standing	 0.679
original_activity_labels_step	 0.5068
original_activity_labels_syncJumping	 0.8992
original_activity_labels_vacuumCleaning	 -0.2189
original_activity_labels_walkingFast	 0.1547
original_activity_labels_walkingNormal	 0.4252
original_activity_labels_walkingSlow	 0.6267
original_activity_labels_walkingStairsUp	 0.1982

coefs = pd.DataFrame(
    slr.coef_, columns=["Coefficients"], index=X_train_std.columns)
coefs

coefs.plot(kind="barh", figsize=(9, 7))
plt.title("MLR model")
plt.axvline(x=0, color=".5")
plt.subplots_adjust(left=0.3)

scores = cross_val_score(estimator=slr,
                         X=X_train_std,
                         y=y_train,
                         cv=10,
                         n_jobs=1)
print('CV keskmine R2 täpsus: %.3f' % np.mean(scores), "+/- %.3f" % np.std(scores))
print('R2 täpsus treeningandmetel: %.3f' % slr.score(X_train_std, y_train))
print('R2 täpsus testandmetel: %.3f' % slr.score(X_test_std, y_test))

CV keskmine R2 täpsus: 0.974 +/- 0.001
R2 täpsus treeningandmetel: 0.974
R2 täpsus testandmetel: 0.974

scores = cross_val_score(estimator=slr,
                         X=X_train_std,
                         y=y_train,
                         scoring = 'neg_mean_squared_error',
                         cv=10,
                         n_jobs=1)
print('CV keskmine RMSE: %.3f' % np.mean(np.sqrt(np.abs(scores))), "+/- %.3f" % np.std(np.sqrt(np.abs(scores))))
print('RMSE treeningandmetel: %.3f' % np.sqrt(mean_squared_error(y_train,slr.predict(X_train_std))))
print('RMSE testandmetel: %.3f' % np.sqrt(mean_squared_error(y_test,slr.predict(X_test_std))))

CV keskmine RMSE: 1.673 +/- 0.035
RMSE treeningandmetel: 1.671
RMSE testandmetel: 1.653

residuals=y_train-slr.predict(X_train_std)

std_residuals=residuals/np.std(residuals)

fig, axs = plt.subplots(2,2,dpi=95,figsize=(15,15))
plt.style.use("seaborn-v0_8-whitegrid")
# Residual against fitted values
axs[0, 0].scatter(x=slr.predict(X_train_std), y=std_residuals)
axs[0, 0].axhline(y=0, color='red', linestyle='dashed')
axs[0, 0].set_xlabel('Fitted Values')
axs[0, 0].set_ylabel('Std. Residuals')
axs[0, 0].set_title('Residuals vs Fitted')

# normal qqplot
stats.probplot(std_residuals, plot=axs[0, 1])
#sm.qqplot(std_residuals, dist=stats.t, fit=True, line='45', c='#4C72B0',ax=axs[0, 1])
axs[0, 1].set_title('Normal Q-Q')

# Fitted values against actual values
axs[1, 0].scatter(x=y_train, y=slr.predict(X_train_std))
axs[1, 0].plot(y_train, y_train, color='red', linestyle='dashed')
axs[1, 0].set_xlabel('Actual Values')
axs[1, 0].set_ylabel('Fitted Values')
axs[1, 0].set_title('Fitted vs Actual')

# Histogram of std. residuals
axs[1, 1].hist(std_residuals, density=True)
x = np.linspace(min(std_residuals),max(std_residuals), 500)
axs[1, 1].plot(x, norm.pdf(x),color='red')
axs[1, 1].set_xlabel('Std. Residuals')
axs[1, 1].set_title('Std. Residuals Density Plot')
fig.tight_layout()
plt.show()

X = df.drop([_SIHTTUNNUS_],axis=1)
y = df[_SIHTTUNNUS_]

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=0)

numerical_columns_selector = selector(dtype_exclude=object)
categorical_columns_selector = selector(dtype_include=object)

numerical_columns = numerical_columns_selector(X)
categorical_columns = categorical_columns_selector(X)

numerical_columns

['EEtot',
 'METS',
 'Rf',
 'VT',
 'VE',
 'O2exp',
 'CO2exp',
 'FeO2',
 'FeCO2',
 'VO2.HR',
 'HR']

categorical_columns

['gender', 'original_activity_labels']

categorical_preprocessor = OneHotEncoder(drop='first')

numerical_preprocessor = StandardScaler()
numerical_preprocessor = Pipeline([
    ('scaler', StandardScaler()),
    ('poly2', PolynomialFeatures(degree=2))
])

preprocessor = ColumnTransformer(
    [
        ("ctg", categorical_preprocessor, categorical_columns),
        ("num", numerical_preprocessor, numerical_columns),
    ]
)

poly_lr = Pipeline([
    ('pre', preprocessor),
    ('lr', LinearRegression())
])

poly_lr.fit(X_train, y_train)

Pipeline(steps=[('pre',
                 ColumnTransformer(transformers=[('ctg',
                                                  OneHotEncoder(drop='first'),
                                                  ['gender',
                                                   'original_activity_labels']),
                                                 ('num',
                                                  Pipeline(steps=[('scaler',
                                                                   StandardScaler()),
                                                                  ('poly2',
                                                                   PolynomialFeatures())]),
                                                  ['EEtot', 'METS', 'Rf', 'VT',
                                                   'VE', 'O2exp', 'CO2exp',
                                                   'FeO2', 'FeCO2', 'VO2.HR',
                                                   'HR'])])),
                ('lr', LinearRegression())])

Pipeline(steps=[('pre',
                 ColumnTransformer(transformers=[('ctg',
                                                  OneHotEncoder(drop='first'),
                                                  ['gender',
                                                   'original_activity_labels']),
                                                 ('num',
                                                  Pipeline(steps=[('scaler',
                                                                   StandardScaler()),
                                                                  ('poly2',
                                                                   PolynomialFeatures())]),
                                                  ['EEtot', 'METS', 'Rf', 'VT',
                                                   'VE', 'O2exp', 'CO2exp',
                                                   'FeO2', 'FeCO2', 'VO2.HR',
                                                   'HR'])])),
                ('lr', LinearRegression())])

ColumnTransformer(transformers=[('ctg', OneHotEncoder(drop='first'),
                                 ['gender', 'original_activity_labels']),
                                ('num',
                                 Pipeline(steps=[('scaler', StandardScaler()),
                                                 ('poly2',
                                                  PolynomialFeatures())]),
                                 ['EEtot', 'METS', 'Rf', 'VT', 'VE', 'O2exp',
                                  'CO2exp', 'FeO2', 'FeCO2', 'VO2.HR', 'HR'])])

['gender', 'original_activity_labels']

OneHotEncoder(drop='first')

['EEtot', 'METS', 'Rf', 'VT', 'VE', 'O2exp', 'CO2exp', 'FeO2', 'FeCO2', 'VO2.HR', 'HR']

StandardScaler()

PolynomialFeatures()

LinearRegression()

preprocessor.get_feature_names_out()

array(['ctg__gender_male', 'ctg__original_activity_labels_dishwashing',
       'ctg__original_activity_labels_lyingDownLeft',
       'ctg__original_activity_labels_lyingDownRight',
       'ctg__original_activity_labels_sittingChair',
       'ctg__original_activity_labels_sittingCouch',
       'ctg__original_activity_labels_sittingSofa',
       'ctg__original_activity_labels_stakingShelves',
       'ctg__original_activity_labels_standing',
       'ctg__original_activity_labels_step',
       'ctg__original_activity_labels_syncJumping',
       'ctg__original_activity_labels_vacuumCleaning',
       'ctg__original_activity_labels_walkingFast',
       'ctg__original_activity_labels_walkingNormal',
       'ctg__original_activity_labels_walkingSlow',
       'ctg__original_activity_labels_walkingStairsUp',
       'ctg__original_activity_labels_nan', 'num__1', 'num__EEtot',
       'num__METS', 'num__Rf', 'num__VT', 'num__VE', 'num__O2exp',
       'num__CO2exp', 'num__FeO2', 'num__FeCO2', 'num__VO2.HR', 'num__HR',
       'num__EEtot^2', 'num__EEtot METS', 'num__EEtot Rf',
       'num__EEtot VT', 'num__EEtot VE', 'num__EEtot O2exp',
       'num__EEtot CO2exp', 'num__EEtot FeO2', 'num__EEtot FeCO2',
       'num__EEtot VO2.HR', 'num__EEtot HR', 'num__METS^2',
       'num__METS Rf', 'num__METS VT', 'num__METS VE', 'num__METS O2exp',
       'num__METS CO2exp', 'num__METS FeO2', 'num__METS FeCO2',
       'num__METS VO2.HR', 'num__METS HR', 'num__Rf^2', 'num__Rf VT',
       'num__Rf VE', 'num__Rf O2exp', 'num__Rf CO2exp', 'num__Rf FeO2',
       'num__Rf FeCO2', 'num__Rf VO2.HR', 'num__Rf HR', 'num__VT^2',
       'num__VT VE', 'num__VT O2exp', 'num__VT CO2exp', 'num__VT FeO2',
       'num__VT FeCO2', 'num__VT VO2.HR', 'num__VT HR', 'num__VE^2',
       'num__VE O2exp', 'num__VE CO2exp', 'num__VE FeO2', 'num__VE FeCO2',
       'num__VE VO2.HR', 'num__VE HR', 'num__O2exp^2',
       'num__O2exp CO2exp', 'num__O2exp FeO2', 'num__O2exp FeCO2',
       'num__O2exp VO2.HR', 'num__O2exp HR', 'num__CO2exp^2',
       'num__CO2exp FeO2', 'num__CO2exp FeCO2', 'num__CO2exp VO2.HR',
       'num__CO2exp HR', 'num__FeO2^2', 'num__FeO2 FeCO2',
       'num__FeO2 VO2.HR', 'num__FeO2 HR', 'num__FeCO2^2',
       'num__FeCO2 VO2.HR', 'num__FeCO2 HR', 'num__VO2.HR^2',
       'num__VO2.HR HR', 'num__HR^2'], dtype=object)

poly_lr.named_steps['lr'].coef_

array([ 1.11588524e+00, -3.35223584e-02, -6.45980021e-02, -1.76905759e-01,
       -8.69381913e-02, -4.17115094e-02, -1.53874216e-01,  1.12142675e-01,
        2.75179062e-01, -1.12167018e-01,  2.49311864e-01, -1.39661167e-03,
       -6.66572688e-02, -2.50763177e-02,  1.42887814e-01, -6.78430960e-02,
       -7.31760935e-02, -2.53504494e+03,  6.31088389e-02, -6.10776529e+00,
        2.70223687e+08, -4.05085758e+08, -1.74074102e+08,  3.44535832e+08,
        2.57149489e+08, -5.28415641e+07, -6.19166777e+07,  7.29776039e+00,
        4.56303008e+00,  1.21282010e-02, -3.55187891e-01, -2.90894779e-02,
       -1.11653483e+00,  5.71564049e-01,  7.31470227e-01,  3.85120749e-01,
       -6.55072629e-02, -3.65646631e-02, -1.36274666e-01, -9.61393714e-02,
       -7.29222342e-01,  1.59336299e-01,  2.72875926e+01,  2.32483765e+00,
       -1.94251935e+01, -6.73374751e+00, -1.92198621e+00, -2.62710661e-01,
       -2.22766016e+00, -1.88767031e-01,  3.10957432e-04,  4.44047038e+06,
        1.36789352e-01, -1.12966873e+09,  1.54127620e+09, -3.34736034e-02,
        5.02254311e-02, -3.12185787e-01, -8.24527442e-02, -3.89861885e+01,
       -3.69538771e+01,  5.26733545e+01,  3.23250163e+01,  3.13555431e+07,
       -2.48821917e+08, -5.35449639e-01, -7.76449083e+00, -1.18688235e+00,
        2.89877477e+01,  7.32735864e+00,  1.30676841e+08, -6.95598797e+08,
        1.13887857e+00, -1.41304341e+00, -1.65728529e+01, -2.49536350e+01,
       -1.22697956e+01,  7.82218382e+08, -3.96077499e-01,  5.49233616e+00,
       -4.65898836e+00, -2.00492064e+08, -2.04347900e-01,  5.53791635e-01,
        2.22513276e+00,  2.94494927e-01,  6.73780438e-01,  6.71497986e-01,
        2.81234711e-01,  2.72435844e-01,  3.55855571e-01, -2.13263795e-01,
        8.28346810e-01,  5.19687173e+00,  4.30181950e-01])

coefs = pd.DataFrame(
    poly_lr.named_steps['lr'].coef_, columns=["Coefficients"], index=preprocessor.get_feature_names_out())
coefs

coefs.plot(kind="barh", figsize=(9, 12))
plt.title("Poly model")
plt.axvline(x=0, color=".5")
plt.subplots_adjust(left=0.3)

scores = cross_val_score(estimator=poly_lr,
                         X=X_train,
                         y=y_train,
                         cv=10,
                         n_jobs=1)
print('CV keskmine R2 täpsus: %.3f' % np.mean(scores), "+/- %.3f" % np.std(scores))
print('Keskmine R2 täpsus treeningandmetel: %.3f' % poly_lr.score(X_train, y_train))
print('Keskmine R2 täpsus testandmetel: %.3f' % poly_lr.score(X_test, y_test))

CV keskmine R2 täpsus: 0.984 +/- 0.001
Keskmine R2 täpsus treeningandmetel: 0.985
Keskmine R2 täpsus testandmetel: 0.984

scores = cross_val_score(estimator=poly_lr,
                         X=X_train,
                         y=y_train,
                         scoring = 'neg_mean_squared_error',
                         cv=10,
                         n_jobs=1)
print('CV keskmine RMSE: %.3f' % np.mean(np.sqrt(np.abs(scores))), "+/- %.3f" % np.std(np.sqrt(np.abs(scores))))
print('RMSE treeningandmetel: %.3f' % np.sqrt(mean_squared_error(y_train,poly_lr.predict(X_train))))
print('RMSE testandmetel: %.3f' % np.sqrt(mean_squared_error(y_test,poly_lr.predict(X_test))))

CV keskmine RMSE: 1.309 +/- 0.029
RMSE treeningandmetel: 1.300
RMSE testandmetel: 1.290

residuals=y_train-poly_lr.predict(X_train)

std_residuals=residuals/np.std(residuals)

fig, axs = plt.subplots(2,2,dpi=95,figsize=(15,15))
# plt.style.use('seaborn')
# Residual against fitted values
axs[0, 0].scatter(x=poly_lr.predict(X_train), y=std_residuals)
axs[0, 0].axhline(y=0, color='red', linestyle='dashed')
axs[0, 0].set_xlabel('Fitted Values')
axs[0, 0].set_ylabel('Std. Residuals')
axs[0, 0].set_title('Residuals vs Fitted')

# normal qqplot
stats.probplot(std_residuals, plot=axs[0, 1])
#sm.qqplot(std_residuals, dist=stats.t, fit=True, line='45', c='#4C72B0',ax=axs[0, 1])
axs[0, 1].set_title('Normal Q-Q')

# Fitted values against actual values
axs[1, 0].scatter(x=y_train, y=poly_lr.predict(X_train))
axs[1, 0].plot(y_train, y_train, color='red', linestyle='dashed')
axs[1, 0].set_xlabel('Actual Values')
axs[1, 0].set_ylabel('Fitted Values')
axs[1, 0].set_title('Fitted vs Actual')

# Histogram of std. residuals
axs[1, 1].hist(std_residuals, density=True)
x = np.linspace(min(std_residuals),max(std_residuals), 500)
axs[1, 1].plot(x, norm.pdf(x),color='red')
axs[1, 1].set_xlabel('Std. Residuals')
axs[1, 1].set_title('Std. Residuals Density Plot')
fig.tight_layout()
plt.show()

degrees = range(1,4)
a = []
c = []
for deg in degrees:
  numerical_preprocessor = Pipeline([
    ('scaler', StandardScaler()),
    ('poly2', PolynomialFeatures(degree=deg))])
  preprocessor = ColumnTransformer(
    [
        ("ctg", categorical_preprocessor, categorical_columns),
        ("num", numerical_preprocessor, numerical_columns),
    ])
  poly_lr = Pipeline([
    ('pre', preprocessor),
    ('lr', LinearRegression())])
  poly_lr.fit(X_train, y_train)
  cv_models = cross_validate(estimator=poly_lr,
                        X=X_train,
                        y=y_train,
                        return_estimator=True,
                        cv=10,
                        n_jobs=1)
  cv_fit = cv_models['estimator']
  c.append(r2_score(np.exp(y_train),np.exp(poly_lr.predict(X_train))))
  b = []
  for i in range(len(cv_fit)):
    b.append(r2_score(np.exp(y_test),np.exp(cv_fit[i].predict(X_test))))
  a.append(np.mean(b))
plt.figure(figsize=(6, 4))

plt.plot(degrees, a, lw=2,
         label='cross-validation test')
plt.plot(degrees, c, lw=2, label='train')

plt.legend(loc='best')
plt.xlabel('degree')
plt.ylabel('R2')
plt.title('Validation curve')
plt.tight_layout()

X = df.drop([_SIHTTUNNUS_],axis=1)
y = df[_SIHTTUNNUS_]

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=0)

numerical_columns_selector = selector(dtype_exclude=object)
categorical_columns_selector = selector(dtype_include=object)

numerical_columns = numerical_columns_selector(X)
categorical_columns = categorical_columns_selector(X)

numerical_columns

['EEtot',
 'METS',
 'Rf',
 'VT',
 'VE',
 'O2exp',
 'CO2exp',
 'FeO2',
 'FeCO2',
 'VO2.HR',
 'HR']

categorical_columns

['gender', 'original_activity_labels']

categorical_preprocessor = OneHotEncoder(drop='first')

numerical_preprocessor = StandardScaler()

preprocessor = ColumnTransformer(
    [
        ("ctg", categorical_preprocessor, categorical_columns),
        ("num", numerical_preprocessor, numerical_columns),
    ]
)
tree_pipe = Pipeline([
    ('pre', preprocessor),
    ('tree', DecisionTreeRegressor(random_state=0))])
tree_pipe.fit(X_train, y_train)

Pipeline(steps=[('pre',
                 ColumnTransformer(transformers=[('ctg',
                                                  OneHotEncoder(drop='first'),
                                                  ['gender',
                                                   'original_activity_labels']),
                                                 ('num', StandardScaler(),
                                                  ['EEtot', 'METS', 'Rf', 'VT',
                                                   'VE', 'O2exp', 'CO2exp',
                                                   'FeO2', 'FeCO2', 'VO2.HR',
                                                   'HR'])])),
                ('tree', DecisionTreeRegressor(random_state=0))])

Pipeline(steps=[('pre',
                 ColumnTransformer(transformers=[('ctg',
                                                  OneHotEncoder(drop='first'),
                                                  ['gender',
                                                   'original_activity_labels']),
                                                 ('num', StandardScaler(),
                                                  ['EEtot', 'METS', 'Rf', 'VT',
                                                   'VE', 'O2exp', 'CO2exp',
                                                   'FeO2', 'FeCO2', 'VO2.HR',
                                                   'HR'])])),
                ('tree', DecisionTreeRegressor(random_state=0))])

ColumnTransformer(transformers=[('ctg', OneHotEncoder(drop='first'),
                                 ['gender', 'original_activity_labels']),
                                ('num', StandardScaler(),
                                 ['EEtot', 'METS', 'Rf', 'VT', 'VE', 'O2exp',
                                  'CO2exp', 'FeO2', 'FeCO2', 'VO2.HR', 'HR'])])

['gender', 'original_activity_labels']

OneHotEncoder(drop='first')

['EEtot', 'METS', 'Rf', 'VT', 'VE', 'O2exp', 'CO2exp', 'FeO2', 'FeCO2', 'VO2.HR', 'HR']

StandardScaler()

DecisionTreeRegressor(random_state=0)

print('Keskmine R2 täpsus treeningandmetel: %.3f' % tree_pipe.score(X_train, y_train))
print('Keskmine R2 täpsus testandmetel: %.3f' % tree_pipe.score(X_test, y_test))

Keskmine R2 täpsus treeningandmetel: 1.000
Keskmine R2 täpsus testandmetel: 0.975

mse = mean_squared_error(y_train, tree_pipe.predict(X_train))
print(f"RMSE testandmetel: {(np.sqrt(mse)):.3f}")

y_pred = tree_pipe.predict(X_test)
mse = mean_squared_error(y_test, y_pred)
rmse = np.sqrt(mse)
print(f"RMSE testandmetel: {(rmse):.3f}")

RMSE testandmetel: 0.000
RMSE testandmetel: 1.623

print('R2 täpsus treeningandmetel: %.3f' % r2_score(np.exp(y_train),np.exp(tree_pipe.predict(X_train))))
print('R2 täpsus testandmetel: %.3f' % r2_score(np.exp(y_test),np.exp(tree_pipe.predict(X_test))))

R2 täpsus treeningandmetel: 1.000
R2 täpsus testandmetel: 0.778

y_pred = np.exp(tree_pipe.predict(X_test))
mse = mean_squared_error(np.exp(y_test), y_pred)
rmse = np.sqrt(mse)
print(f"RMSE testandmetel: {(rmse):.3f}")

RMSE testandmetel: 123540818094081495006943483507626838327296.000

imp = pd.DataFrame(tree_pipe[1].feature_importances_)
ft = pd.DataFrame(preprocessor.get_feature_names_out())
ft_imp = pd.concat([ft,imp],axis=1)
ft_imp.columns = ['Feature', 'Importance']
ft_imp.sort_values(by='Importance',ascending=False)

from sklearn.model_selection import GridSearchCV

parameters={"tree__splitter":["best","random"],
            "tree__max_depth" : [1,3,5,7,9],
           "tree__min_samples_leaf":[1,2,3,4,5,6,7],
           "tree__max_features":["log2","sqrt",None],
           "tree__max_leaf_nodes":[None,10,20,30] }

gs_tree_pipe = GridSearchCV(estimator=tree_pipe, param_grid=parameters, cv=5, verbose=0)

gs_tree_pipe.fit(X_train, y_train)

gs_tree_pipe.best_params_

{'tree__max_depth': 9,
 'tree__max_features': None,
 'tree__max_leaf_nodes': None,
 'tree__min_samples_leaf': 7,
 'tree__splitter': 'best'}

print(f"R2 score on train: {gs_tree_pipe.score(X_train, y_train):.3f}")
print(f"R2 score on train: {gs_tree_pipe.score(X_test, y_test):.3f}")

R2 score on train: 0.984
R2 score on train: 0.978

mse = mean_squared_error(y_train, gs_tree_pipe.predict(X_train))
print(f"RMSE testandmetel: {(np.sqrt(mse)):.3f}")

y_pred = gs_tree_pipe.predict(X_test)
mse = mean_squared_error(y_test, y_pred)
rmse = np.sqrt(mse)
print(f"RMSE testandmetel: {(rmse):.3f}")

RMSE testandmetel: 1.330
RMSE testandmetel: 1.537

print(f"R2 score on train: {r2_score(np.exp(y_train),np.exp(gs_tree_pipe.predict(X_train))):.3f}")
print(f"R2 score on test: {r2_score(np.exp(y_test),np.exp(gs_tree_pipe.predict(X_test))):.3f}")

R2 score on train: 0.942
R2 score on test: 0.908

y_pred = np.exp(gs_tree_pipe.predict(X_test))
mse = mean_squared_error(np.exp(y_test), y_pred)
rmse = np.sqrt(mse)
print(f"RMSE testandmetel: {(rmse):.3f}")

RMSE testandmetel: 79616768612770933240036264162054455164928.000

rf_pipe = Pipeline([
    ('pre', preprocessor),
    ('rf', RandomForestRegressor(random_state=0))])

rf_pipe.fit(X_train, y_train)

Pipeline(steps=[('pre',
                 ColumnTransformer(transformers=[('ctg',
                                                  OneHotEncoder(drop='first'),
                                                  ['gender',
                                                   'original_activity_labels']),
                                                 ('num', StandardScaler(),
                                                  ['EEtot', 'METS', 'Rf', 'VT',
                                                   'VE', 'O2exp', 'CO2exp',
                                                   'FeO2', 'FeCO2', 'VO2.HR',
                                                   'HR'])])),
                ('rf', RandomForestRegressor(random_state=0))])

Pipeline(steps=[('pre',
                 ColumnTransformer(transformers=[('ctg',
                                                  OneHotEncoder(drop='first'),
                                                  ['gender',
                                                   'original_activity_labels']),
                                                 ('num', StandardScaler(),
                                                  ['EEtot', 'METS', 'Rf', 'VT',
                                                   'VE', 'O2exp', 'CO2exp',
                                                   'FeO2', 'FeCO2', 'VO2.HR',
                                                   'HR'])])),
                ('rf', RandomForestRegressor(random_state=0))])

ColumnTransformer(transformers=[('ctg', OneHotEncoder(drop='first'),
                                 ['gender', 'original_activity_labels']),
                                ('num', StandardScaler(),
                                 ['EEtot', 'METS', 'Rf', 'VT', 'VE', 'O2exp',
                                  'CO2exp', 'FeO2', 'FeCO2', 'VO2.HR', 'HR'])])

['gender', 'original_activity_labels']

OneHotEncoder(drop='first')

['EEtot', 'METS', 'Rf', 'VT', 'VE', 'O2exp', 'CO2exp', 'FeO2', 'FeCO2', 'VO2.HR', 'HR']

StandardScaler()

RandomForestRegressor(random_state=0)

print('Keskmine R2 täpsus treeningandmetel: %.3f' % rf_pipe.score(X_train, y_train))
print('Keskmine R2 täpsus testandmetel: %.3f' % rf_pipe.score(X_test, y_test))

Keskmine R2 täpsus treeningandmetel: 0.998
Keskmine R2 täpsus testandmetel: 0.989

mse = mean_squared_error(y_train, rf_pipe.predict(X_train))
print(f"Random Forest mudeli RMSE testandmetel: {(np.sqrt(mse)):.3f}")

y_pred = rf_pipe.predict(X_test)
mse = mean_squared_error(y_test, y_pred)
rmse = np.sqrt(mse)
print(f"Random Forest mudeli RMSE testandmetel: {(rmse):.3f}")

Random Forest mudeli RMSE testandmetel: 0.457
Random Forest mudeli RMSE testandmetel: 1.191

print('R2 täpsus treeningandmetel: %.3f' % r2_score(np.exp(y_train),np.exp(rf_pipe.predict(X_train))))
print('R2 täpsus testandmetel: %.3f' % r2_score(np.exp(y_test),np.exp(rf_pipe.predict(X_test))))

R2 täpsus treeningandmetel: 0.987
R2 täpsus testandmetel: 0.896

imp = pd.DataFrame(rf_pipe[1].feature_importances_)
ft = pd.DataFrame(preprocessor.get_feature_names_out())
ft_imp = pd.concat([ft,imp],axis=1)
ft_imp.columns = ['Feature', 'Importance']
ft_imp.sort_values(by='Importance',ascending=False)

# param_grid_rf = {
#     'rf__n_estimators': [10, 50, 100, 500, 1000],
#     'rf__max_features': ['log2', 'sqrt', 0.8,1]
# }

param_grid_rf = {
    'rf__n_estimators': [10, 50, 100],
    'rf__max_features': ['log2', 'sqrt', 0.8,1]
}

gs_rf_pipe = GridSearchCV(estimator=rf_pipe, param_grid=param_grid_rf, cv=5, verbose=0)

gs_rf_pipe.fit(X_train, y_train)

gs_rf_pipe.best_params_

{'rf__max_features': 0.8, 'rf__n_estimators': 100}

print(gs_rf_pipe.score(X_train, y_train))
print(gs_rf_pipe.score(X_test, y_test))

0.9985355652170951
0.9894740695731525

mse = mean_squared_error(y_train, gs_rf_pipe.predict(X_train))
print(f"Random Forest mudeli RMSE testandmetel: {(np.sqrt(mse)):.3f}")

y_pred = gs_rf_pipe.predict(X_test)
mse = mean_squared_error(y_test, y_pred)
rmse = np.sqrt(mse)
print(f"Random Forest mudeli RMSE testandmetel: {(rmse):.3f}")

Random Forest mudeli RMSE testandmetel: 0.400
Random Forest mudeli RMSE testandmetel: 1.058

print(f"R2 score on train: {r2_score(np.exp(y_train),np.exp(gs_rf_pipe.predict(X_train))):.3f}")
print(f"R2 score on test: {r2_score(np.exp(y_test),np.exp(gs_rf_pipe.predict(X_test))):.3f}")

R2 score on train: 0.986
R2 score on test: 0.910

y_pred = np.exp(gs_rf_pipe.predict(X_test))
mse = mean_squared_error(np.exp(y_test), y_pred)
rmse = np.sqrt(mse)
print(f"Random Forest mudeli RMSE testandmetel: {(rmse):.3f}")

Random Forest mudeli RMSE testandmetel: 256069566736770598900636151744551404437504.000

preprocessor.get_feature_names_out()

array(['ctg__gender_male', 'ctg__original_activity_labels_dishwashing',
       'ctg__original_activity_labels_lyingDownLeft',
       'ctg__original_activity_labels_lyingDownRight',
       'ctg__original_activity_labels_sittingChair',
       'ctg__original_activity_labels_sittingCouch',
       'ctg__original_activity_labels_sittingSofa',
       'ctg__original_activity_labels_stakingShelves',
       'ctg__original_activity_labels_standing',
       'ctg__original_activity_labels_step',
       'ctg__original_activity_labels_syncJumping',
       'ctg__original_activity_labels_vacuumCleaning',
       'ctg__original_activity_labels_walkingFast',
       'ctg__original_activity_labels_walkingNormal',
       'ctg__original_activity_labels_walkingSlow',
       'ctg__original_activity_labels_walkingStairsUp',
       'ctg__original_activity_labels_nan', 'num__EEtot', 'num__METS',
       'num__Rf', 'num__VT', 'num__VE', 'num__O2exp', 'num__CO2exp',
       'num__FeO2', 'num__FeCO2', 'num__VO2.HR', 'num__HR'], dtype=object)

# rf_pipe = Pipeline([
#     ('pre', preprocessor),
#     ('rf',RandomForestRegressor(max_features='log2', n_estimators= 1000))])

rf_pipe = Pipeline([
    ('pre', preprocessor),
    ('rf',RandomForestRegressor(max_features='log2', n_estimators= 100))])
rf_pipe.fit(X_train, y_train)

Pipeline(steps=[('pre',
                 ColumnTransformer(transformers=[('ctg',
                                                  OneHotEncoder(drop='first'),
                                                  ['gender',
                                                   'original_activity_labels']),
                                                 ('num', StandardScaler(),
                                                  ['EEtot', 'METS', 'Rf', 'VT',
                                                   'VE', 'O2exp', 'CO2exp',
                                                   'FeO2', 'FeCO2', 'VO2.HR',
                                                   'HR'])])),
                ('rf', RandomForestRegressor(max_features='log2'))])

Pipeline(steps=[('pre',
                 ColumnTransformer(transformers=[('ctg',
                                                  OneHotEncoder(drop='first'),
                                                  ['gender',
                                                   'original_activity_labels']),
                                                 ('num', StandardScaler(),
                                                  ['EEtot', 'METS', 'Rf', 'VT',
                                                   'VE', 'O2exp', 'CO2exp',
                                                   'FeO2', 'FeCO2', 'VO2.HR',
                                                   'HR'])])),
                ('rf', RandomForestRegressor(max_features='log2'))])

ColumnTransformer(transformers=[('ctg', OneHotEncoder(drop='first'),
                                 ['gender', 'original_activity_labels']),
                                ('num', StandardScaler(),
                                 ['EEtot', 'METS', 'Rf', 'VT', 'VE', 'O2exp',
                                  'CO2exp', 'FeO2', 'FeCO2', 'VO2.HR', 'HR'])])

['gender', 'original_activity_labels']

OneHotEncoder(drop='first')

['EEtot', 'METS', 'Rf', 'VT', 'VE', 'O2exp', 'CO2exp', 'FeO2', 'FeCO2', 'VO2.HR', 'HR']

StandardScaler()

RandomForestRegressor(max_features='log2')

imp = pd.DataFrame(rf_pipe[1].feature_importances_)
ft = pd.DataFrame(preprocessor.get_feature_names_out())
ft_imp = pd.concat([ft,imp],axis=1)
ft_imp.columns = ['Feature', 'Importance']
ft_imp.sort_values(by='Importance',ascending=False)

	count	mean	std	min	25%	50%	75%	max
EEtot	38695.00	73.08	52.11	0.00	34.37	63.59	100.59	291.44
METS	38695.00	2.85	1.81	0.00	1.47	2.39	3.87	15.23
Rf	38695.00	23.99	10.80	2.89	18.13	22.30	27.78	375.00
BR	38695.00	82.02	10.42	23.00	77.00	85.00	90.00	99.00
VT	38695.00	1.19	0.60	0.04	0.76	1.08	1.50	4.64
VE	38695.00	27.73	16.97	0.20	15.40	22.84	35.15	132.13
O2exp	38695.00	207.24	103.41	7.55	134.52	186.63	257.73	969.38
CO2exp	38695.00	39.00	23.83	0.00	21.42	33.96	51.19	167.76
FeO2	38695.00	17.49	0.75	12.74	17.05	17.51	17.94	22.43
FeCO2	38695.00	3.11	0.68	0.00	2.70	3.09	3.53	6.11
VO2.HR	38695.00	7.94	4.99	0.00	4.60	7.55	11.12	39.39
HR	38695.00	81.68	36.18	0.00	69.00	83.00	104.00	203.00

	count	mean	std	min	25%	50%	75%	max
EEtot	30956.0000	-0.0000	1.0000	-1.4077	-0.7433	-0.1782	0.5320	4.1906
METS	30956.0000	0.0000	1.0000	-1.5789	-0.7663	-0.2625	0.5628	6.8422
Rf	30956.0000	-0.0000	1.0000	-1.9714	-0.5403	-0.1544	0.3578	32.8557
VT	30956.0000	-0.0000	1.0000	-1.9071	-0.7096	-0.1964	0.5062	5.6983
VE	30956.0000	-0.0000	1.0000	-1.6217	-0.7266	-0.2861	0.4337	6.1408
O2exp	30956.0000	0.0000	1.0000	-1.9263	-0.7007	-0.2029	0.4874	7.3464
CO2exp	30956.0000	0.0000	1.0000	-1.6377	-0.7359	-0.2141	0.5081	5.3928
FeO2	30956.0000	-0.0000	1.0000	-6.2942	-0.5822	0.0304	0.5923	6.5488
FeCO2	30956.0000	0.0000	1.0000	-4.5911	-0.6006	-0.0221	0.6297	4.4391
VO2.HR	30956.0000	0.0000	1.0000	-1.5906	-0.6696	-0.0786	0.6387	5.4592
HR	30956.0000	-0.0000	1.0000	-2.2608	-0.3523	0.0350	0.6158	3.3541

	EEtot	METS	Rf	VT	VE	O2exp	CO2exp	FeO2	FeCO2	VO2.HR	HR
7985	120.9925	1.0065	9.9834	1.2768	12.7465	232.7153	31.6134	18.2268	2.4760	4.6235	64
16281	144.2626	3.4369	31.4136	0.9985	31.3650	174.7038	35.1855	17.4974	3.5240	6.5937	135
9594	103.5986	1.5816	18.2927	0.9607	17.5744	167.2611	29.5767	17.4098	3.0786	7.8152	68
9242	51.8238	2.3240	26.6667	0.9985	26.6258	174.7062	30.6985	17.4974	3.0746	11.8315	66
17936	129.4891	7.6673	32.2581	1.6683	53.8148	286.4388	65.4659	17.1699	3.9242	0.0000	0

	EEtot	METS	Rf	VT	VE	O2exp	CO2exp	FeO2	FeCO2	VO2.HR	HR
14331	60.2944	1.5900	22.3048	0.6506	14.5125	113.0974	16.8269	17.3824	2.5862	6.3275	73
3717	118.9672	3.1380	28.3019	1.1517	32.5956	197.4944	43.9670	17.1479	3.8175	10.5206	95
6326	46.2270	1.5039	22.1402	0.7110	15.7408	126.2795	19.9925	17.7618	2.8121	5.6145	75
33591	99.7225	1.6865	18.8679	0.8608	16.2412	153.3907	22.7434	17.8199	2.6422	5.6727	77
9741	129.1196	1.5356	20.6897	0.9097	18.8221	161.0398	27.0269	17.7018	2.9709	7.2668	71

	count	mean	std	min	25%	50%	75%	max
EEtot	30956.0000	-0.0000	1.0000	-1.4077	-0.7433	-0.1782	0.5320	4.1906
METS	30956.0000	0.0000	1.0000	-1.5789	-0.7663	-0.2625	0.5628	6.8422
Rf	30956.0000	-0.0000	1.0000	-1.9714	-0.5403	-0.1544	0.3578	32.8557
VT	30956.0000	-0.0000	1.0000	-1.9071	-0.7096	-0.1964	0.5062	5.6983
VE	30956.0000	-0.0000	1.0000	-1.6217	-0.7266	-0.2861	0.4337	6.1408
O2exp	30956.0000	0.0000	1.0000	-1.9263	-0.7007	-0.2029	0.4874	7.3464
CO2exp	30956.0000	0.0000	1.0000	-1.6377	-0.7359	-0.2141	0.5081	5.3928
FeO2	30956.0000	-0.0000	1.0000	-6.2942	-0.5822	0.0304	0.5923	6.5488
FeCO2	30956.0000	0.0000	1.0000	-4.5911	-0.6006	-0.0221	0.6297	4.4391
VO2.HR	30956.0000	0.0000	1.0000	-1.5906	-0.6696	-0.0786	0.6387	5.4592
HR	30956.0000	-0.0000	1.0000	-2.2608	-0.3523	0.0350	0.6158	3.3541

	ID	trial_date	gender	age	weight	height	bmi	gaAnkle	gaChest	gaWrist	...	R	Ti	Te	Ttot	VO2.HR	HR	Qt	SV	original_activity_labels	predicted_activity_label
0	GOTOV05	08/02/2016 13:42	female	61.6000	68.6000	162	26.1000	True	True	True	...	0.9846	0.9300	1.8600	2.7900	2.1250	102	0	0	NaN	sitting
1	GOTOV05	08/02/2016 13:42	female	61.6000	68.6000	162	26.1000	True	True	True	...	1.0035	1.2600	1.1800	2.4400	2.2403	103	0	0	NaN	NaN
2	GOTOV05	08/02/2016 13:42	female	61.6000	68.6000	162	26.1000	True	True	True	...	1.0399	0.9700	1.6900	2.6600	4.2051	104	0	0	NaN	NaN
3	GOTOV05	08/02/2016 13:42	female	61.6000	68.6000	162	26.1000	True	True	True	...	1.0635	0.9600	2.0400	3.0000	4.3329	106	0	0	lyingDownRight	standing
4	GOTOV05	08/02/2016 13:42	female	61.6000	68.6000	162	26.1000	True	True	True	...	1.0307	1.1500	1.5700	2.7200	2.6086	106	0	0	lyingDownRight	NaN
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
38718	GOTOV36	30/05/2016 10:41	female	81.3000	72.0000	167	25.8000	True	True	True	...	0.8145	1.0100	1.7400	2.7500	3.9375	77	0	0	NaN	standing
38719	GOTOV36	30/05/2016 10:41	female	81.3000	72.0000	167	25.8000	True	True	True	...	0.7833	1.1100	1.9100	3.0200	3.1757	77	0	0	NaN	NaN
38720	GOTOV36	30/05/2016 10:41	female	81.3000	72.0000	167	25.8000	True	True	True	...	0.7643	0.8300	1.4100	2.2400	7.3279	77	0	0	NaN	NaN
38721	GOTOV36	30/05/2016 10:41	female	81.3000	72.0000	167	25.8000	True	True	True	...	0.7948	1.2500	2.7700	4.0200	4.4365	77	0	0	NaN	NaN
38722	GOTOV36	30/05/2016 10:41	female	81.3000	72.0000	167	25.8000	True	True	True	...	0.7068	1.1300	1.2900	2.4200	4.7681	77	0	0	NaN	NaN

	gender	EEtot	METS	Rf	BR	VT	VE	O2exp	CO2exp	FeO2	FeCO2	VO2.HR	HR	original_activity_labels
0	female	0.0000	0.9528	21.5054	92	0.4091	8.7978	73.1476	12.4463	17.8803	3.0424	2.1250	102	NaN
1	female	0.0097	1.0143	24.5902	90	0.4642	11.4144	85.4919	11.8342	18.4176	2.5495	2.2403	103	NaN
2	female	0.0940	1.9223	22.5564	84	0.8774	19.7902	159.3537	25.3007	18.1628	2.8837	4.2051	104	NaN
3	female	0.2080	2.0189	20.0000	85	0.9243	18.4858	164.5567	30.6057	17.8035	3.3113	4.3329	106	lyingDownRight
4	female	0.2703	1.2154	22.0588	89	0.5876	12.9624	107.3240	16.2210	18.2639	2.7604	2.6086	106	lyingDownRight
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
38718	female	85.4218	1.2031	21.8182	90	0.5650	12.3281	102.0947	13.8710	18.0686	2.4549	3.9375	77	NaN
38719	female	85.4871	0.9703	19.8675	92	0.5314	10.5573	96.9951	11.8312	18.2534	2.2265	3.1757	77	NaN
38720	female	85.5828	2.2391	26.7857	87	0.6436	17.2386	110.4581	19.6846	17.1632	3.0586	7.3279	77	NaN
38721	female	85.7260	1.3556	14.9254	92	0.7017	10.4733	120.3514	22.2344	17.1512	3.1686	4.4365	77	NaN
38722	female	86.1708	1.4569	24.7934	89	0.6028	14.9449	109.3362	12.8511	18.1387	2.1320	4.7681	77	NaN

	count	unique	top	freq
gender	38695	2	male	24786
original_activity_labels	24432	16	cycling	4212

	EEtot	METS	Rf	BR	VT	VE	O2exp	CO2exp	FeO2	FeCO2	VO2.HR	HR
EEtot	1.00	0.46	0.18	-0.51	0.43	0.55	0.44	0.40	-0.02	0.16	0.35	0.32
METS	0.46	1.00	0.26	-0.91	0.73	0.90	0.71	0.77	-0.44	0.56	0.69	0.39
Rf	0.18	0.26	1.00	-0.35	-0.14	0.32	-0.13	-0.15	0.23	-0.19	0.10	0.14
BR	-0.51	-0.91	-0.35	1.00	-0.73	-0.97	-0.74	-0.67	0.14	-0.31	-0.63	-0.47
VT	0.43	0.73	-0.14	-0.73	1.00	0.79	1.00	0.95	-0.32	0.47	0.68	0.44
VE	0.55	0.90	0.32	-0.97	0.79	1.00	0.80	0.73	-0.15	0.31	0.71	0.50
O2exp	0.44	0.71	-0.13	-0.74	1.00	0.80	1.00	0.92	-0.25	0.41	0.67	0.44
CO2exp	0.40	0.77	-0.15	-0.67	0.95	0.73	0.92	1.00	-0.51	0.67	0.68	0.39
FeO2	-0.02	-0.44	0.23	0.14	-0.32	-0.15	-0.25	-0.51	1.00	-0.91	-0.37	-0.01
FeCO2	0.16	0.56	-0.19	-0.31	0.47	0.31	0.41	0.67	-0.91	1.00	0.42	0.12
VO2.HR	0.35	0.69	0.10	-0.63	0.68	0.71	0.67	0.68	-0.37	0.42	1.00	0.58
HR	0.32	0.39	0.14	-0.47	0.44	0.50	0.44	0.39	-0.01	0.12	0.58	1.00

	Coefficients
EEtot	0.2114
METS	-4.8676
Rf	-0.1695
VT	4.9437
VE	-6.6501
O2exp	-7.2140
CO2exp	3.1740
FeO2	-0.7719
FeCO2	-1.1161
VO2.HR	1.1979
HR	-0.4380

	EEtot	METS	Rf	VT	VE	O2exp	CO2exp	FeO2	FeCO2	VO2.HR	...	original_activity_labels_sittingSofa	original_activity_labels_stakingShelves	original_activity_labels_standing	original_activity_labels_step	original_activity_labels_syncJumping	original_activity_labels_vacuumCleaning	original_activity_labels_walkingFast	original_activity_labels_walkingNormal	original_activity_labels_walkingSlow	original_activity_labels_walkingStairsUp
0	0.9165	-1.0224	-1.3076	0.1345	-0.8833	0.2415	-0.3129	0.9779	-0.9337	-0.6636	...	False	False	False	False	False	True	False	False	False	False
1	1.3635	0.3213	0.6981	-0.3264	0.2122	-0.3180	-0.1632	0.0106	0.6142	-0.2686	...	False	False	False	False	False	False	False	False	False	False
2	0.5823	-0.7044	-0.5299	-0.3889	-0.5992	-0.3898	-0.3982	-0.1057	-0.0438	-0.0237	...	False	False	False	False	False	False	False	False	False	False
3	-0.4122	-0.2940	0.2538	-0.3264	-0.0667	-0.3180	-0.3512	0.0106	-0.0497	0.7816	...	False	False	False	False	False	False	False	False	False	False
4	1.0797	2.6602	0.7772	0.7827	1.5331	0.7596	1.1058	-0.4238	1.2053	-1.5906	...	False	False	False	False	False	False	False	False	False	False

	EEtot	METS	Rf	VT	VE	O2exp	CO2exp	FeO2	FeCO2	VO2.HR	...	original_activity_labels_sittingSofa	original_activity_labels_stakingShelves	original_activity_labels_standing	original_activity_labels_step	original_activity_labels_syncJumping	original_activity_labels_vacuumCleaning	original_activity_labels_walkingFast	original_activity_labels_walkingNormal	original_activity_labels_walkingSlow	original_activity_labels_walkingStairsUp
0	-0.2495	-0.6998	-0.1544	-0.9023	-0.7794	-0.9122	-0.9325	-0.1420	-0.7710	-0.3220	...	False	False	False	False	False	False	False	False	False	False
1	0.8776	0.1560	0.4069	-0.0726	0.2846	-0.0982	0.2048	-0.4530	1.0478	0.5188	...	False	False	False	False	False	False	False	False	False	False
2	-0.5197	-0.7474	-0.1698	-0.8025	-0.7071	-0.7851	-0.7999	0.3612	-0.4374	-0.4649	...	False	False	False	False	False	False	False	False	False	False
3	0.5079	-0.6465	-0.4761	-0.5544	-0.6777	-0.5236	-0.6846	0.4382	-0.6883	-0.4532	...	False	False	False	False	False	False	False	False	False	False
4	1.0726	-0.7299	-0.3056	-0.4733	-0.5258	-0.4498	-0.5051	0.2816	-0.2029	-0.1336	...	False	False	False	False	False	True	False	False	False	False

	Coefficients
EEtot	0.3412
METS	-3.5980
Rf	-0.1433
VT	-1.9545
VE	-7.7387
O2exp	-1.0557
CO2exp	3.7042
FeO2	-0.9021
FeCO2	-1.1345
VO2.HR	1.0252
HR	-0.4310
gender_male	1.6887
original_activity_labels_dishwashing	-0.0310
original_activity_labels_lyingDownLeft	-0.0819
original_activity_labels_lyingDownRight	0.1211
original_activity_labels_sittingChair	0.1296
original_activity_labels_sittingCouch	0.2812
original_activity_labels_sittingSofa	0.3217
original_activity_labels_stakingShelves	-0.0936
original_activity_labels_standing	0.6790
original_activity_labels_step	0.5068
original_activity_labels_syncJumping	0.8992
original_activity_labels_vacuumCleaning	-0.2189
original_activity_labels_walkingFast	0.1547
original_activity_labels_walkingNormal	0.4252
original_activity_labels_walkingSlow	0.6267
original_activity_labels_walkingStairsUp	0.1982

	Coefficients
ctg__gender_male	1.1159
ctg__original_activity_labels_dishwashing	-0.0335
ctg__original_activity_labels_lyingDownLeft	-0.0646
ctg__original_activity_labels_lyingDownRight	-0.1769
ctg__original_activity_labels_sittingChair	-0.0869
...	...
num__FeCO2 VO2.HR	0.3559
num__FeCO2 HR	-0.2133
num__VO2.HR^2	0.8283
num__VO2.HR HR	5.1969
num__HR^2	0.4302

	Feature	Importance
21	num__VE	0.9461
0	ctg__gender_male	0.0187
27	num__HR	0.0094
18	num__METS	0.0062
26	num__VO2.HR	0.0040
25	num__FeCO2	0.0039
23	num__CO2exp	0.0025
24	num__FeO2	0.0024
17	num__EEtot	0.0022
19	num__Rf	0.0021
22	num__O2exp	0.0009
20	num__VT	0.0006
16	ctg__original_activity_labels_nan	0.0002
2	ctg__original_activity_labels_lyingDownLeft	0.0001
12	ctg__original_activity_labels_walkingFast	0.0001
13	ctg__original_activity_labels_walkingNormal	0.0001
11	ctg__original_activity_labels_vacuumCleaning	0.0001
14	ctg__original_activity_labels_walkingSlow	0.0001
8	ctg__original_activity_labels_standing	0.0000
7	ctg__original_activity_labels_stakingShelves	0.0000
3	ctg__original_activity_labels_lyingDownRight	0.0000
1	ctg__original_activity_labels_dishwashing	0.0000
6	ctg__original_activity_labels_sittingSofa	0.0000
4	ctg__original_activity_labels_sittingChair	0.0000
9	ctg__original_activity_labels_step	0.0000
5	ctg__original_activity_labels_sittingCouch	0.0000
15	ctg__original_activity_labels_walkingStairsUp	0.0000
10	ctg__original_activity_labels_syncJumping	0.0000

	Feature	Importance
21	num__VE	0.2356
18	num__METS	0.1878
22	num__O2exp	0.1071
20	num__VT	0.0865
19	num__Rf	0.0792
26	num__VO2.HR	0.0774
27	num__HR	0.0670
23	num__CO2exp	0.0621
17	num__EEtot	0.0364
24	num__FeO2	0.0198
25	num__FeCO2	0.0174
0	ctg__gender_male	0.0077
12	ctg__original_activity_labels_walkingFast	0.0047
16	ctg__original_activity_labels_nan	0.0027
5	ctg__original_activity_labels_sittingCouch	0.0016
13	ctg__original_activity_labels_walkingNormal	0.0015
3	ctg__original_activity_labels_lyingDownRight	0.0012
11	ctg__original_activity_labels_vacuumCleaning	0.0008
6	ctg__original_activity_labels_sittingSofa	0.0007
4	ctg__original_activity_labels_sittingChair	0.0007
14	ctg__original_activity_labels_walkingSlow	0.0006
8	ctg__original_activity_labels_standing	0.0005
1	ctg__original_activity_labels_dishwashing	0.0004
2	ctg__original_activity_labels_lyingDownLeft	0.0003
7	ctg__original_activity_labels_stakingShelves	0.0003
9	ctg__original_activity_labels_step	0.0001
15	ctg__original_activity_labels_walkingStairsUp	0.0000
10	ctg__original_activity_labels_syncJumping	0.0000

Meetod	R2 train	R2 test	RMSE train	RMSE test
Lineaarregressiooni mudel ainult arvuliste tunnustega	0.970	0.970	1.807	1.786
Lineaarregressiooni mudel koos kategooriliste tunnustega	0.974	0.974	1.671	1.653
Polünomiaalregressioon	0.985	0.984	1.309	1.300
Otsustuspuu regressioon	1.000	0.975	0.000	1.623
Otsustuspuu regressioon GridSearchCV tuunimisel	0.984	0.978	1.330	1.537
Random Forest regressioon	0.998	0.989	0.457	1.191
Random Forest regressioon GridSearchCV tuunimisel	0.998	0.989	0.400	1.058

Inimse hingamise sageduse ja teiste füüsiliste parameetrite regressioonanalüüs¶

ITB8814 Andmekaevandamine, Projekt¶

Autor: Juri Lunin¶

Sissejuhatus¶

Andmestiku kirjeldus¶

Uurimiseesmärk¶

Tööülesanded (tööhüpoteesid)¶

Metoodika ja uuringu käik¶

Sisukord¶

Andmeanalüüs¶

Andmete lugemine¶

Ülevaade andmestiku struktuurist¶

Andmestiku puhastamine¶

Duplikaatide kontroll¶

Tunnuste teisendamine¶

Puuduvate andmetega objektide kontroll¶

Tunnuste kirjeldus¶

Arvuliste tunnuste kirjeldus¶

Mittearvuliste atribuutide kirjeldus¶

Sihttunnuse Y kirjeldus¶

Seoste analüüs¶

Seosed arvuliste tunnuste vahel¶

Korrelatsioonimaatriks¶

Sihttunnuse seos mittearvuliste tunnustega¶

Lineaarregressioon¶

Mudeli konstrueerimine: arvuliste prediktoritega mudel (ilma kategoriaalsete tunnusteta)¶

Andmete ettevalmistamine¶

Treening- ja testandmete eraldamine¶

Andmete standardiseerimine¶

Mudeli loomine¶

Mudeli testimine¶

Lineaarregressiooni mudel kategooriliste tunnustega¶

Polünomiaalregressioon¶

Andmete ettevalmistamine¶

Mudeli loomine treeningandmetel¶

Polünomiaalregressiooni astme validatsioonikõver¶

Otsustuspuu regressioon¶

Andmete ettevalmistamine¶

Otsustuspuu regressioonimudel¶

Numbriliste ja kategooriliste muutujate mudelis kasutamine.¶

Random Forest regressioon¶

Tulemuste analüüs¶

Kokkuvõte¶